nichenetpy

NicheNetPy: the python implementation of the NicheNet method. (ported from NicheNetR) The goal of NicheNet is to study intercellular communication from a computational perspective. NicheNet uses human or mouse gene expression data of interacting cells as input and combines this with a prior model that integrates existing knowledge on ligand-to-target signaling paths. This allows to predict ligand-receptor interactions that might drive gene expression changes in cells of interest.

We describe the NicheNet algorithm in the following paper: NicheNet: modeling intercellular communication by linking ligands to target genes.

Installation of nichenetpy

pip install nichenetpy

Overview of NicheNet

Background

NicheNet strongly differs from most computational approaches to study cell-cell communication (CCC), as summarized conceptually by the figure below (top panel: current ligand-receptor inference approaches; bottom panel: NicheNet). Many approaches to study CCC from expression data involve linking ligands expressed by sender cells to their corresponding receptors expressed by receiver cells. However, functional understanding of a CCC process also requires knowing how these inferred ligand-receptor interactions result in changes in the expression of downstream target genes within the receiver cells. Therefore, we developed NicheNet to consider the gene regulatory effects of ligands.



At the core of NicheNet is a prior knowledge model, created by integrating three types of databases—ligand-receptor interactions, signaling pathways, and transcription factor (TF) regulation—to form a complete communication network spanning from ligands to their downstream target genes (see figure below). Therefore, this model goes beyond ligand-receptor interactions and incorporates intracellular signaling and transcriptional regulation as well. As a result, NicheNet is able to predict which ligands influence the expression in another cell, which target genes are affected by each ligand, and which signaling mediators may be involved. By generating these novel types of hypotheses, NicheNet can drive an improved functional understanding of a CCC process of interest. We provide a pre-built prior model, it is also possible to construct your own model (see notebooks below).

Main functionalities of nichenetpy

  • Assessing how well ligands expressed by a sender cell can predict changes in gene expression in the receiver cell
  • Prioritizing ligands based on their effect on gene expression
  • Inferring putative ligand-target links active in the system under study
  • Inferring potential signaling paths between ligands and target genes of interest: to generate causal hypotheses and check which data sources support the predictions
  • Construction of user-defined prior ligand-target models

Moreover, we provide instructions on how to make intuitive visualizations of the main predictions (e.g., via circos plots as shown here below).



As input to NicheNet, users must provide cell type-annotated expression data that reflects a cell-cell communication (CCC) event. The input can be single-cell or sorted bulk data from human or mouse. As output, NicheNet returns the ranking of ligands that best explain the CCC event of interest, as well as candidate target genes with high potential to be regulated by these ligands. As an intermediate step, we extract the three features required for the analysis: a list of potential ligands, a gene set that captures the downstream effects of the CCC event of interest, and a background set of genes. Further explanation on each feature can be found in the introductory notebooks.

Learning to use nichenetpy

A very basic tutorial for people who are unfamiliar with python can be found here.

The following notebooks contain the explanation on how to perform a basic NicheNet analysis on an AnnData object. This includes prioritizing ligands and predicting target genes of prioritized ligands. We recommend starting with the step-by-step analysis, but we also demonstrate the use of a single wrapper function.

Case study on HNSCC tumor which demonstrates the flexibility of NicheNet. Here, the gene set of interest was determined by the original authors, and the expression data is a matrix rather than an AnnData object.

The following notebooks explain how to do some follow-up analyses:

If you want to make a circos plot visualization of the NicheNet output to show active ligand-target links between interacting cells, you can check following notebooks:

People interested in building their own models or benchmarking their own models against NicheNet can read the following notebooks:

For a comparison between Seurat's FindAllMarkers (which is ported into nichenetpy) and Scanpy's rank_genes_groups, see the following notebook:

FAQ

References

Browaeys, R., Saelens, W. & Saeys, Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods (2019)

Bonnardel et al. Stellate Cells, Hepatocytes, and Endothelial Cells Imprint the Kupffer Cell Identity on Monocytes Colonizing the Liver Macrophage Niche. Immunity (2019)

Guilliams et al. Spatial proteogenomics reveals distinct and evolutionarily conserved hepatic macrophage niches. Cell (2022) 

  1'''
  2**NicheNetPy: the python implementation of the NicheNet method. (ported from [NicheNetR](https://github.com/saeyslab/nichenetr/tree/master))** The goal of
  3NicheNet is to study intercellular communication from a computational
  4perspective. NicheNet uses human or mouse gene expression data of
  5interacting cells as input and combines this with a prior model that
  6integrates existing knowledge on ligand-to-target signaling paths. This
  7allows to predict ligand-receptor interactions that might drive gene
  8expression changes in cells of interest.
  9
 10We describe the NicheNet algorithm in the following paper: [NicheNet:
 11modeling intercellular communication by linking ligands to target
 12genes](https://www.nature.com/articles/s41592-019-0667-5).
 13
 14## Installation of nichenetpy
 15
 16```
 17pip install nichenetpy
 18```
 19
 20## Overview of NicheNet
 21
 22### Background
 23
 24NicheNet strongly differs from most computational approaches to study
 25cell-cell communication (CCC), as summarized conceptually by the figure
 26below (**top panel:** current ligand-receptor inference approaches;
 27**bottom panel:** NicheNet). Many approaches to study CCC from
 28expression data involve linking ligands expressed by sender cells to
 29their corresponding receptors expressed by receiver cells. However,
 30functional understanding of a CCC process also requires knowing how
 31these inferred ligand-receptor interactions result in changes in the
 32expression of downstream target genes within the receiver cells.
 33Therefore, we developed NicheNet to consider the gene regulatory effects
 34of ligands. <br><br>
 35<img src="https://github.com/saeyslab/nichenetpy/raw/main/images/comparison_other_approaches_2.jpg"
 36width="450" /> <br><br>
 37
 38At the core of NicheNet is a prior knowledge model, created by
 39integrating three types of databases—ligand-receptor interactions,
 40signaling pathways, and transcription factor (TF) regulation—to form a
 41complete communication network spanning from ligands to their downstream
 42target genes (see figure below). Therefore, this model goes beyond
 43ligand-receptor interactions and incorporates intracellular signaling
 44and transcriptional regulation as well. As a result, NicheNet is able to
 45predict which ligands influence the expression in another cell, which
 46target genes are affected by each ligand, and which signaling mediators
 47may be involved. By generating these novel types of hypotheses, NicheNet
 48can drive an improved functional understanding of a CCC process of
 49interest. We provide a pre-built prior model, it is
 50also possible to construct your own model (see notebooks below).
 51
 52<img src="https://github.com/saeyslab/nichenetpy/raw/main/images/nichenet_prior_model.png"
 53style="width:70.0%" />
 54
 55### Main functionalities of nichenetpy
 56
 57-   Assessing how well ligands expressed by a sender cell can predict
 58    changes in gene expression in the receiver cell
 59-   Prioritizing ligands based on their effect on gene expression
 60-   Inferring putative ligand-target links active in the system under
 61    study
 62-   Inferring potential signaling paths between ligands and target genes
 63    of interest: to generate causal hypotheses and check which data
 64    sources support the predictions
 65-   Construction of user-defined prior ligand-target models
 66
 67Moreover, we provide instructions on how to make intuitive
 68visualizations of the main predictions (e.g., via circos plots as shown
 69here below).
 70
 71<br><br>
 72<img src="https://github.com/saeyslab/nichenetpy/blob/main/images/circos.png" width="600" />
 73
 74As input to NicheNet, users must provide cell type-annotated expression
 75data that reflects a cell-cell communication (CCC) event. The input can
 76be single-cell or sorted bulk data from human or mouse. As output,
 77NicheNet returns the ranking of ligands that best explain the CCC event
 78of interest, as well as candidate target genes with high potential to be
 79regulated by these ligands. As an intermediate step, we extract the
 80three features required for the analysis: a list of potential ligands, a
 81gene set that captures the downstream effects of the CCC event of
 82interest, and a background set of genes. Further explanation on each
 83feature can be found in the introductory notebooks.
 84
 85<img src="https://github.com/saeyslab/nichenetpy/raw/main/images/figure1.svg" width="600" />
 86
 87## Learning to use nichenetpy
 88
 89A very basic tutorial for people who are unfamiliar with python can be found here. 
 90
 91-   [Basic Tutorial](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/basic_tutorial.ipynb)
 92
 93The following notebooks contain the explanation on how to perform a
 94basic NicheNet analysis on an AnnData object. This includes prioritizing
 95ligands and predicting target genes of prioritized ligands. We recommend
 96starting with the step-by-step analysis, but we also demonstrate the use
 97of a single wrapper function. 
 98
 99-   [Perform NicheNet analysis starting from an AnnData object:
100    step-by-step
101    analysis](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/steps.ipynb)
102-   [Perform NicheNet analysis starting from an AnnData
103    object: wrapper](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/wrapper.ipynb)
104
105Case study on HNSCC tumor which demonstrates the flexibility of
106NicheNet. Here, the gene set of interest was determined by the original
107authors, and the expression data is a matrix rather than an AnnData
108object.
109
110-   [NicheNet’s ligand activity analysis on a gene set of
111    interest](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/ligand_activity_geneset.ipynb)
112
113The following notebooks explain how to do some follow-up
114analyses:
115
116-   [Prioritization of ligands based on expression
117    values](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/steps_prioritization.ipynb)
118-   [Inferring ligand-to-target signaling
119    paths](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/ligand_target_signaling_path.ipynb)
120-   [Assess how well top-ranked ligands can predict a gene set of
121    interest](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/target_prediction_evaluation_geneset.ipynb)
122-   [Single-cell NicheNet’s ligand activity
123    analysis](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/ligand_activity_single_cell.ipynb)
124
125If you want to make a circos plot visualization of the NicheNet output
126to show active ligand-target links between interacting cells, you can
127check following notebooks:
128
129-   [circos visualization](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/circos.ipynb)
130
131People interested in building their own models or benchmarking their own
132models against NicheNet can read the following notebooks:
133
134-   [Model construction](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/model_construction.ipynb)
135-   [Using LIANA ligand-receptor databases to construct the ligand-target model](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/model_construction_with_liana.ipynb)
136-   [Model evaluation: target gene and ligand activity prediction](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/model_evaluation.ipynb)
137
138For a comparison between Seurat's FindAllMarkers (which is ported into nichenetpy) and Scanpy's rank_genes_groups, see the following notebook:
139
140-   [Comparison: Seurat vs Scanpy](https://github.com/saeyslab/nichenetpy/blob/main/notebooks/seuratVSscanpy.ipynb)
141
142## FAQ
143
144-   Check the FAQ pages at [FAQ NicheNetPy](https://github.com/saeyslab/nichenetpy/blob/main/faq.md) and [FAQ Nichenet](https://github.com/saeyslab/nichenetr/blob/master/vignettes/faq.md)
145
146## References
147
148Browaeys, R., Saelens, W. & Saeys, Y. NicheNet: modeling intercellular
149communication by linking ligands to target genes. Nat Methods (2019)
150<doi:10.1038/s41592-019-0667-5>
151
152Bonnardel et al. Stellate Cells, Hepatocytes, and Endothelial Cells
153Imprint the Kupffer Cell Identity on Monocytes Colonizing the Liver
154Macrophage Niche. Immunity (2019) <doi:10.1016/j.immuni.2019.08.017>
155
156Guilliams et al. Spatial proteogenomics reveals distinct and
157evolutionarily conserved hepatic macrophage niches. Cell (2022)
158<doi:10.1016/j.cell.2021.12.018>
159'''
160
161__version__ = "1.0.1"